Short-term power load prediction method based on multilayer neural network

Abstract

The invention discloses a short-term power load prediction method based on a multilayer neural network. The method comprises the following steps of: normalizing a power consumption value and an air temperature value in a power load data set containing a plurality of continuous moments with the same time interval, the power consumption of each moment and the air temperature of each moment, and then representing a power load characteristic matrix and a power load true value vector; constructing a network structure of a short-term power load prediction model based on a multilayer neural network by using the matrix; sequentially dividing a data set expressed as the matrix and the vector into a training set, a verification set and a test set according to a ratio of 7: 2: 1; performing parameter adjustment on the prediction model by using a training set and a verification set to obtain an optimized model SPLF-MNN; using the test set to calculate the prediction precision of the SPLF-MNN model so as to evaluate the performance of the model, and using the SPLF-MNN model to predict the power load at a plurality of moments in the next day or week. The SPLF-MNN model considers the influence of the air temperature on the power load, the prediction precision is high, the training speed is high, and the application prospect is wide.

Description

technical field [0001] The invention belongs to the technical field of short-term power load forecasting, and relates to a short-term power load forecasting method based on a neural network, in particular to a short-term power load forecasting method based on a multilayer neural network. Background technique [0002] Power load forecasting is an important part of power system planning and the basis of power system economic operation. The basis of power load forecasting is power load data. Such data is a kind of time series data, which usually contains multiple consecutive moments with the same time interval (moment contains date and time), power load at each moment (such as power consumption amount), meteorological data (such as temperature) at each moment, etc. A certain amount of electric load data constitutes an electric load data set. Power load forecasting can dig out the law of power load changes based on time-related historical data such as power loads and weather, ...

AI Technical Summary

Problems solved by technology

[0018] To sum up, in order to overcome the above-mentioned deficiencies in the prior art, namely: the short-term power load forecasting model does not consider the impact of temperature factors on the power load, the prediction accuracy is not high enough, and the training speed of the model is slow, the present invention uses the attention mechanism Combined with the convolutional neural network to form a specific multi-layer neural network, thus providing a short-term power load forecasting method based on multi-layer neural network (SPLF for short) -MNN)

Images